On February 22nd, Alibaba launched the "Medical AI Multi Cancer Early Screening Public Welfare Project" in Lishui, Zhejiang. This project aims to apply cutting-edge medical AI technology innovation from Damo Hospital to the field of healthcare, aiming to achieve early screening of multiple cancers through large-scale random daily examinations and improve the local digital health level. It is also the first project in the country to achieve early screening of multiple cancers through AI, which means that cutting-edge AI research has gone out of the laboratory and is expected to be promoted to more regions with uneven medical resources.
The PANDA deep learning model developed by the medical AI team of Dharma Hospital is the first to build a large-scale screening method for early pancreatic cancer, mainly through the method of "plain CT+AI".
Combining plain CT with AI technology to assist in early screening of multiple cancer types
It is reported that the project relies on the self-developed intelligent film reading product "Da Yi Zhi Ying" by the medical AI laboratory of Alibaba Da Mo Hospital. Through routine chest and abdominal CT plain scans, it helps diagnosis and treatment through a breakthrough "plain CT+AI" method.
When explaining the innovation of medical AI technology in medicine to the Daily Economic News reporter, Lu Chenying, the director of the Radiology Department and Nuclear Medicine Department of Lishui Central Hospital, introduced that doctors usually need to carefully examine hundreds of medical image slices of each patient during chest CT plain scan. "From the left lung to the right lung, and then to the mediastinum, ribs, and hilum, it needs to be viewed layer by layer.", This process is not only time-consuming but also prone to causing fatigue for doctors, thereby affecting the accuracy and efficiency of diagnosis.
By combining imaging and AI in cancer screening, doctors have greatly improved their film reading efficiency: from manually identifying a single disease in 5-15 minutes to identifying multiple diseases in 2-3 minutes. In addition, in terms of accuracy, previous results published on Nature Medicine showed that the PANDA deep learning model had an accuracy rate of 92.9% in identifying lesions and 99.9% in identifying disease-free conditions in over 200000 population verifications; 31 cases of clinical misdiagnosis were found and 2 cases were cured.
The doctor is conducting disease screening and diagnosis through Dayi Zhiying. Photo taken by journalist Xu Libo [align]
Lu Chenying said that in addition to the diagnosis of lung diseases, Alibaba's medical AI can also be used to examine a variety of other organs, such as the pancreas and other upper abdominal organs. Diseases in these parts, such as pancreatic cancer, are often difficult to find in traditional chest CT plain scanning. In addition, Lu Chenying emphasized that the main objective of chest CT is to examine lung lesions, while the upper abdomen, such as the liver and pancreas, scanned simultaneously, is usually considered as an incidental examination and not the main target. "If the lesion in the upper abdomen is very obvious, the doctor will record it; if it is not obvious, it may not be recorded because the main focus is on the lungs."
Data shows that one of the reasons for the difficulty in early detection of pancreatic cancer is that there are few obvious symptoms and it is difficult to conduct voluntary examinations before the cancer progresses. In addition, due to the fact that the pancreas is located at the deepest part of the body, some routine imaging examinations may not be able to display the entire pancreas. However, enhanced CT, nuclear magnetic resonance and other imaging diagnosis are not suitable for large-scale screening of pancreatic cancer due to the need for injection of contrast agent, radiation dose, long inspection cycle, high cost and other reasons.
Therefore, if medical AI can be applied, pancreatic cancer can be preliminarily screened on the simplest and most conventional plain CT, which will have important clinical significance for early screening and early treatment of pancreatic cancer. In other words, the main value of the "Medical AI Multi cancer Early Screening Public Welfare Project" launched by Alibaba in Lishui this time is to endow simple and low-cost plain scan CT with the ability to screen pancreatic cancer, so that while improving the detection rate, it will not bring extra radiation and economic burden to patients, and ultimately greatly improve the coverage of pancreatic cancer screening.
Alibaba disclosed that the project will start with early screening of pancreatic cancer and osteoporosis in Lishui, and gradually access the screening capacity of liver cancer, esophageal cancer, gastric cancer, colon cancer, fatty liver and other cancers and chronic diseases.
The challenge of compatibility still needs to be addressed from theory to practice
In this project landing in Lishui, Alibaba's medical AI technology mainly focuses on medical imaging diagnosis, especially the analysis of CT plain scan images.
According to Guo Jianfei, a product expert at Alibaba Damo Hospital's medical AI laboratory, theoretically, this AI technology does not rely on specific devices and can be applied to any standard CT image. However, translating this technology from theory to practice, especially in hospitals at different levels, still faces many challenges.
Guo Jianfei stated that there are significant differences in the standardization level of technology and equipment among hospitals at different levels, from grassroots hospitals to Lishui Central Hospital, and then to top tier tertiary hospitals in China. This difference is not only reflected in the level of equipment, but also in data processing and management. Alibaba's medical AI project team realizes that in order to successfully implement this technology, it is necessary to work closely with local hospitals, continuously adapt, and jointly solve data level problems, including data collection, processing, and analysis.
Zhou Yongjin, the head of the Abdominal Group of the Radiology Department at Lishui Central Hospital, also revealed to reporters that initially, there were compatibility issues in the data transmitted by the hospital to the AI system, which resulted in the AI model being unable to accurately identify lesions. After continuous technical debugging, AI began to be able to more accurately identify and analyze cases, especially showing high recognition rates in the detection of certain lesions.
Through cooperation with local hospitals such as Lishui Central Hospital, medical AI projects can be more closely aligned with practical medical scenarios, gradually optimizing and adjusting their algorithms to make them more precise and efficient. Alibaba also stated that the "medical AI multi cancer early screening" Lishui model, jointly developed by multiple parties, will be further promoted to hospitals in other cities across the country after maturity, achieving inclusiveness in the medical field and benefiting more people.
Lu Chenying said that in addition to the diagnosis of lung diseases, Alibaba's medical AI can also be used to examine a variety of other organs, such as the pancreas and other upper abdominal organs. Diseases in these parts, such as pancreatic cancer, are often difficult to find in traditional chest CT plain scanning. In addition, Lu Chenying emphasized that the main objective of chest CT is to examine lung lesions, while the upper abdomen, such as the liver and pancreas, scanned simultaneously, is usually considered as an incidental examination and not the main target. "If the lesion in the upper abdomen is very obvious, the doctor will record it; if it is not obvious, it may not be recorded because the main focus is on the lungs."
Data shows that one of the reasons for the difficulty in early detection of pancreatic cancer is that there are few obvious symptoms and it is difficult to conduct voluntary examinations before the cancer progresses. In addition, due to the fact that the pancreas is located at the deepest part of the body, some routine imaging examinations may not be able to display the entire pancreas. However, enhanced CT, nuclear magnetic resonance and other imaging diagnosis are not suitable for large-scale screening of pancreatic cancer due to the need for injection of contrast agent, radiation dose, long inspection cycle, high cost and other reasons.
Therefore, if medical AI can be applied, pancreatic cancer can be preliminarily screened on the simplest and most conventional plain CT, which will have important clinical significance for early screening and early treatment of pancreatic cancer. In other words, the main value of the "Medical AI Multi cancer Early Screening Public Welfare Project" launched by Alibaba in Lishui this time is to endow simple and low-cost plain scan CT with the ability to screen pancreatic cancer, so that while improving the detection rate, it will not bring extra radiation and economic burden to patients, and ultimately greatly improve the coverage of pancreatic cancer screening.
Alibaba disclosed that the project will start with early screening of pancreatic cancer and osteoporosis in Lishui, and gradually access the screening capacity of liver cancer, esophageal cancer, gastric cancer, colon cancer, fatty liver and other cancers and chronic diseases.
The challenge of compatibility still needs to be addressed from theory to practice
In this project landing in Lishui, Alibaba's medical AI technology mainly focuses on medical imaging diagnosis, especially the analysis of CT plain scan images.
According to Guo Jianfei, a product expert at Alibaba Damo Hospital's medical AI laboratory, theoretically, this AI technology does not rely on specific devices and can be applied to any standard CT image. However, translating this technology from theory to practice, especially in hospitals at different levels, still faces many challenges.
Guo Jianfei stated that there are significant differences in the standardization level of technology and equipment among hospitals at different levels, from grassroots hospitals to Lishui Central Hospital, and then to top tier tertiary hospitals in China. This difference is not only reflected in the level of equipment, but also in data processing and management. Alibaba's medical AI project team realizes that in order to successfully implement this technology, it is necessary to work closely with local hospitals, continuously adapt, and jointly solve data level problems, including data collection, processing, and analysis.
Zhou Yongjin, the head of the Abdominal Group of the Radiology Department at Lishui Central Hospital, also revealed to reporters that initially, there were compatibility issues in the data transmitted by the hospital to the AI system, which resulted in the AI model being unable to accurately identify lesions. After continuous technical debugging, AI began to be able to more accurately identify and analyze cases, especially showing high recognition rates in the detection of certain lesions.
Through cooperation with local hospitals such as Lishui Central Hospital, medical AI projects can be more closely aligned with practical medical scenarios, gradually optimizing and adjusting their algorithms to make them more precise and efficient. Alibaba also stated that the "medical AI multi cancer early screening" Lishui model, jointly developed by multiple parties, will be further promoted to hospitals in other cities across the country after maturity, achieving inclusiveness in the medical field and benefiting more people.
转播
